Types of Data

Nominal = unordered categories (blood group, sex). Ordinal = ordered categories without guaranteed equal spacing (pain score, triage category). Interval / ratio = continuous numbers (blood pressure, age) — ratio data also has a true zero. Nominal and ordinal are categorical; interval and ratio are continuous.

The four data types ranked by how much information each carries. Nominal labels merely name a category; ordinal adds a rank order; interval adds equal, measurable gaps; ratio adds a meaningful zero so ratios make sense (40 kg really is twice 20 kg).

You can always step down the ladder (treat age as an age-band) but never up — you can’t invent equal spacing that isn’t there. Categorical data (nominal/ordinal) → proportions and χ² tests; continuous data (interval/ratio) → mean or median with a t-test or Mann–Whitney.

Pick the wrong rung and you pick the wrong test. Averaging an ordinal 1–5 pain score as if the gaps were equal, or running a t-test on category counts, produces numbers that look precise but mean nothing — a favourite trap in appraisal questions.

• Nominal / Ordinal → proportions, χ²
• Interval / Ratio → mean/median, t-test / Mann–Whitney
• Only ratio data has a true zero

Everyday ED example — a pain VAS recorded 0–10 and a triage category (1–5) are both ordinal: a patient with a “6” isn’t in twice as much pain as a “3”, and the gap between triage 1 and 2 isn’t the same as between 4 and 5. So these are summarised with the median (IQR) and compared with non-parametric tests — not the mean and a t-test. Reporting a “mean pain score of 4.7” treats ordinal data as interval — statistically improper, however common it is in practice.

• Ordinal-as-interval — averaging ranked scores whose gaps aren’t guaranteed equal.
• Choosing the wrong test for the data type (e.g. a t-test on categorical counts instead of χ²).
• Confusing discrete count data (number of attendances) with truly continuous data.